By Bob Haavind, Editorial Director, Solid State Technology
A detailed economic model suggests that industry productivity will lag historical trends early in the next decade, slowing the decline in cost/transistor, unless new initiatives get underway to reverse this trend, reported Scott Kramer, director, International SEMATECH Manufacturing Initiative (ISMI), in a Tuesday afternoon presentation at The ConFab.
Many lessons learned during the transition to 300mm wafers may help the industry find productivity improvements needed to maintain a healthy industry growth curve, he suggested. ISMI is hoping to apply this learning to improve next-generation fabs and to develop strategies for the next wafer size transition.
Kramer showed that historically the compound annual growth rate (CAGR) trend for cost/transistor has been about -29%. But over the 2005-2021 time period, using only 300mm or smaller wafers, this is projected to decline to about -22%, using an ISMI economic model (SEE CHART).
There are a number of options for boosting productivity, he pointed out, including improved overall equipment effectiveness (OEE), accelerated technology innovation, a larger wafer size, and perhaps new design approaches that could reduce the number of mask levels. He then projected how much some of these improvements might be expected to contribute, assuming that any shift to 450mm wafers was held off until 2012 or beyond. Even with a 20% compounded improvement in OEE, and accelerated technology to maintain two-year cycles for shifts to 65nm and 45nm, the effects would still fall short of Moore’s Law, he indicated.
Kramer also showed that starting a transition to 450mm wafers in 2011/2012 could greatly reduce the number of fabs needed to provide sufficient capacity to meet projected growing demand for chips. With 300mm only, he showed that the ISMI model projected that more than 500 200mm fabs with 20k wafer starts/month would be needed by the 2018 time frame, vs. only about half as many with a combination of 300mm and 450mm wafer processing.
Admittedly, any wafer size transition will require industry consensus and global coordination, Kramer pointed out. Consensus on an industry strategy for a transition would have to be driven by rigorous analysis of economic, business, and technology impacts. Even then, any plans would have to be continuously evaluated and adjusted, including any impact from changing technology and market dynamics that could influence transition timing. In the manufacturing strategy, he said, there would have to be a comprehensive look at productivity improvements, bridging, and backward applications.
To get all of this started, it would be vital to establish early agreement on initial, fundamental decisions and on common global chip manufacturing requirements. Standards should be developed early, he suggested, and the standards cycle might be accelerated by prototyping of early standards, along with validation and concurrent conformance test development.
Kramer noted that ISMI member company guidelines would be needed for productivity requirements, factory architecture options and vision, and for factory operations.
An ISMI/SEMI joint productivity working group could address an economic assessment and develop industry strategies. This collaboration could lead to mutual understanding of economic and technical challenges while developing options for addressing them.
Industry summits could help maintain communication and help the industry to coalesce on a direction, Kramer believes. A SEMI Manufacturing Technology Forum could develop recommendations for standards initiatives.
ISMI could contribute to the global effort by doing economic modeling to support analysis of cost and value curves, he suggested. An ISMI factory architecture simulation could evaluate productivity options and define visions for fab architectures.
In addition, an ISMI/SEMI joint productivity working group could enable supplier engagement to validate modeling assumptions and to reflect industry reality, Kramer suggested.
Continuous productivity improvement is taking place in 300mm fabs, and ISMI proposes a “300 Prime” (300